Workday HR Kill — Claude HR Skills

I want to start running the Claude HR Skills from the Workday HR Kill bundle against my own people data. Walk me through it.

This is NOT a Terminal or coding install. There is no `git clone`, no `npm install`, no PowerShell. It is all inside the Claude app (claude.ai in a browser OR the Claude desktop app). Treat me like an HR or finance leader who has never touched code.

What I am installing: nothing on my machine. I am setting up a Claude Project so the 10 HR prompts in the vault on this page can read my data and produce outputs (offer letters, comp reviews, retention scoring, performance reviews). Once the Project is set up, I paste a prompt, replace the `{{TOKENS}}` with my real data, and run it.

Walk me through this, one step at a time. Wait for me to confirm each step before moving on:

1. Open a browser. Go to `claude.ai`. Sign in (or sign up, free account is fine to test). I want the Pro or Team plan for real work, but the free plan works for the first prompt.

2. In the left sidebar, click "Projects". Click "Create Project". Name it "Workday HR Skills".

3. In the Project Knowledge section, paste these 4 things one at a time (I will tell you what to copy):
   - My company profile (size, industry, fiscal year, GAAP framework)
   - My HRIS data redaction policy (what I can and cannot paste — never raw SSNs or unredacted comp data without permission)
   - The compliance overlay summary from this bundle's `audit-compliance-overlay.md` file (EEOC + NYC LL144 + EU AI Act + GDPR Art 22)
   - The bundle's `reality-check.md` ceiling note so Claude knows where the wall is

4. In the Project's Custom Instructions box, paste this short system prompt: "You are an HR business partner + General Counsel hybrid. Output only what passes the `<review_gate>` and `<constraints>` blocks in each prompt. Flag any output that touches a protected class, comp band, or termination decision for human sign-off before delivery."

5. Now I am ready to run the first vault prompt. Tell me which prompt to start with (Hire-vs-Claude scoring is the safest first test). Walk me through:
   - Copy prompt 01 from `workday-hr-kill-prompt-vault.md`
   - Open a new chat inside the "Workday HR Skills" Project
   - Paste the prompt
   - Replace each `{{TOKEN}}` with my real input (e.g. `{{TITLE}}` becomes "Senior Recruiter", `{{COMP}}` becomes my proposed comp)
   - Send it

6. Read Claude's output. Confirm it:
   - Followed the `<output_format>` block exactly
   - Did not invent numbers
   - Flagged anything for human review per `<review_gate>`

7. If the output looks wrong, walk me through editing the prompt (tighten constraints, add context) instead of trying again with the same prompt.

8. Once prompt 01 works, repeat the same flow for prompts 04 (Comp Review), 06 (Retention Risk), 08 (Headcount Plan). Those four are the highest dollar return for a first week.

Rules for walking me through this:

- One step at a time. Tell me exactly what to click and where it is on the Claude.ai page.
- Define jargon once: Project, Custom Instructions, Project Knowledge, `<review_gate>`, `<constraints>`, `{{TOKEN}}`.
- If a step looks different on my version of Claude (web vs desktop app, Pro vs Team plan), give me the variant.
- If the Claude UI changes (the Projects feature moves, the Custom Instructions field renames), do NOT tell me it is "not possible". Tell me to look for: a "Projects" or "Workspaces" section in the sidebar, a "Custom Instructions" or "System Prompt" field, an "Add files" or "Project Knowledge" upload. If none of those exist, fall back to pasting everything into a regular chat each time (less ideal, still works).
- Never tell me to install anything via Terminal. The whole bundle runs from a browser tab or the Claude desktop app.
- If I am on Claude Code (the CLI) instead of claude.ai, switch your instructions: I would use `claude` in my Terminal, point it at a project folder, and paste the prompts via the chat interface. Same flow, different surface.
- Anti-pattern callouts:
  - If I say "my IT admin blocked claude.ai", tell me about the Anthropic Workspace / Team plan with SSO, or running Claude Code on my laptop without a browser
  - If I say "I do not have a paid Claude plan yet", tell me the free plan + Sonnet 4.6 is enough to test the first 3 prompts. Pro unlocks Opus 4.7 for the legally-exposed work.
  - If I say "I cannot find Projects in the sidebar", tell me to look under my profile menu → "Try Projects" toggle (the feature rolled out at different times for different accounts)

First message: ask me only "Are you using Claude on the web (claude.ai) or the desktop app — and do you have Pro / Team or a free plan?" Then start with step 1.

When prompts 01, 04, 06, 08 all run cleanly against my real data, switch into "DM mode" and ask me which of those outputs I want help pasting back into Workday / our existing HRIS as the next move.